The Department of Energy is one of the few federal funders that writes checks directly to for-profit businesses — but there’s an important catch for AI companies about what it will and won’t fund. Here’s the honest picture for 2026, cited to official sources. (dgm builds AI; proposal writing is separate — see the end.)

DOE SBIR/STTR — direct funding for businesses

DOE runs its own SBIR/STTR program, and unlike many federal grants, it’s aimed squarely at companies: participation is limited to for-profit US small businesses that meet SBA eligibility. Awards follow the government-wide structure — Phase I for feasibility (up to about $323,090) and Phase II for development (up to about $2,153,927) — and are non-dilutive grants or cooperative agreements.

The honest limit on AI eligibility

Here’s what many “DOE AI grant” pages won’t tell you: DOE’s funding topics center on energy, advanced manufacturing, quantum information science, semiconductors, and critical materials — and AI is not a standalone topic on the program’s landing page. That means an AI project generally has to tie to a DOE mission — AI for the power grid, for materials discovery, for scientific computing — to be eligible. A generic “we want to adopt AI in our business” project does not fit.

So DOE is a strong option if your AI work intersects energy or science, and the wrong door if it doesn’t. Be honest with yourself about which you are before investing in an application.

ARPA-E — high-risk energy R&D

ARPA-E funds high-risk, high-reward energy technology, and for-profit entities can apply directly as lead applicants or team members through Funding Opportunity Announcements. Award sizes vary by FOA (often low millions). The same rule applies: AI is fundable only as a tool inside an energy project — not as a standalone AI grant. Live opportunities are posted on the ARPA-E eXCHANGE portal.

When DOE is — and isn’t — the right door

DOE/ARPA-E make sense if your AI advances an energy or scientific mission and you’re doing genuine R&D. If your AI work is general business automation with no energy or science angle, you’ll get further with NSF’s AI-focused SBIR or, if you’re building/customizing software, the R&D tax credit and Section 174A expensing.

How dgm helps

dgm implements osFoundry and builds AI for US businesses. We help develop the underlying technology a DOE or ARPA-E project would fund, while the proposal itself is best handled with an SBIR/STTR advisor or the DOE program office. If DOE isn’t the right fit for your AI work, we’ll tell you that too.